nanocodec_decoder.mlmodelc: re-export with compute_units=CPU_AND_NE + FP32 precision (Kokoro ANE recipe). Strictly better than prior export — clean audio on both ANE and GPU; the prior FP16 export had high-frequency hiss when run on ANE.

Add files using upload-large-folder tool